Worked on the apache/celeborn repository to enhance Hadoop FileSystem shutdown handling, focusing on improving data integrity and stability for distributed systems using S3 storage. Addressed a critical bug by implementing a patch that prevents the ShutdownHookManager from prematurely closing Hadoop FileSystems, ensuring all streams are properly closed before shutdown. This solution reduces the risk of incomplete files and errors when accessing shuffle data, particularly benefiting long-running jobs and cloud-based workloads. Utilized Scala to deliver this fix, applying expertise in file systems and Hadoop to increase reliability for both streaming and batch pipelines without introducing new features during the development period.